1

IEEE 802.3-2018 §40.8.2 Crossover function reads:

Although the automatic MDI/MDI-X configuration (see 40.4.4) is not required for successful operation of 1000BASE-T, a crossover function must be implemented for every link segment to support the operation of Auto-Negotiation. The crossover function connects the transmitters of one PHY to the receivers of the PHY at the other end of the link segment. For a PHY that does not implement the crossover function, the MDI labels in the middle column of Table 40–12 refer to its own internal circuits. For PHYs that do implement the internal crossover, the MDI labels in the last column of Table 40–12 refer to the internal circuits of the remote PHY of the link segment.

For reference, here's Table 40–12:

Table 40–12—Assignment of PMA signal to MDI and MDI-X pin-outs

Breaking §40.8.2 down sentence by sentence, I have questions:

Although the automatic MDI/MDI-X configuration (see 40.4.4) is not required for successful operation of 1000BASE-T, a crossover function must be implemented for every link segment to support the operation of Auto-Negotiation.

Auto-Negotiation "builds upon the link pulse mechanism employed by 10BASE-T" (§28.2.1.1), and 10BASE-T specifies a crossover function that "connects the transmitter of one MAU to the receiver of the MAU at the other end of the twisted-pair link" (§14.5.2). So it's clear that Auto-Negotiation requires a crossover function.

But Auto-Negotiation also specifies a Transmit Switch function that enables "the transmit path from a single technology-dependent PMA to the MDI once a highest common denominator choice has been made and Auto-Negotiation has completed" (§28.2.1.3).

Q: Why doesn't the Auto-Negotiation Transmit Switch function insulate 1000BASE-T from having to define a crossover function? Isn't the 1000BASE-T PMA disconnected while Auto-Negotiation is active?

The crossover function connects the transmitters of one PHY to the receivers of the PHY at the other end of the link segment.

This sentence echoes the definition of the 10BASE-T crossover function, which "connects the transmitter of one MAU to the receiver of the MAU at the other end of the twisted-pair link" (§14.5.2). Such a sentence makes sense in the context of 10BASE-T, which transmits on one pair and receives on another. But it doesn't make sense in the context of 1000BASE-T, which simultaneously transmits and receives on all four pairs. Any wiring (crossover or straight) results in connecting "the transmitters" to "the receivers".

Q: Is this sentence vestigial?

For a PHY that does not implement the crossover function, the MDI labels in the middle column of Table 40–12 refer to its own internal circuits. For PHYs that do implement the internal crossover, the MDI labels in the last column of Table 40–12 refer to the internal circuits of the remote PHY of the link segment.

Consider the common case of a link composed of one MDI PHY and one MDI-X PHY connected with a straight-through cable. If we interpret the first sentence literally, then contact 1 is assigned to PMA signal BI_DA+ in the MDI PHY. But if we interpret the second sentence literally, then contact 1 is assigned to PMA signal BI_DB+ in the remote (i.e., MDI) PHY.

Q: How can one contact be assigned to two different PMA signals in the same PHY?

Q: Is it true that the BI_DA signal pair in one PHY is connected to the BI_DA signal pair in the link partner PHY?

Improve this question

1 Answer 1

Reset to default
1

Auto Negotiation takes place before any PHY can link. AN requires a crossover to be in place (e.g. MDI to MDI-X with straight cable, MDI to MDI with crossover cable) unless Auto MDI-X is used an at least one side.

Why doesn't the Auto-Negotiation Transmit Switch function insulate 1000BASE-T from having to define a crossover function? Isn't the 1000BASE-T PMA disconnected while Auto-Negotiation is active?

1000BASE-T requires Auto Negotiation to work. AN uses fast link pulses (FLP) that are not part of any specific PHY but shared for all of them. Therefore, AN must use the same method with any port, regardless of which PHY variants it supports. Yes, the 1000BASE-T PMA (or any other) is inactive while AN is running.

Is this sentence vestigial?

I don't think so. It says that pair 1-2 on MDI is called BI_DA, and on MDI-X it connects to BI_DB on the remote MDI, which is on pair 3-6. That is the crossover function exactly.

How can one contact be assigned to two different PMA signals in the same PHY?

It isn't. The names are the ones for the local pins on MDI and for the remote pins on MDI-X.

Is it true that the BI_DA signal pair in one PHY is connected to the BI_DA signal pair in the link partner PHY?

No, that won't work.

IEEE 802.3 may have a weird way of putting it but knowing their process a bit, it's likely the result of a long debate. You need to read very carefully, but then it's unambiguous.

Improve this answer
3
  • "The names are the ones for the local pins on MDI and for the remote pins on MDI-X." Well, that's what §40.8.2 says, but it doesn't make sense to me. If the MDI column is referring to the local (MDI) PHY, as it says, and the MDI-X column is referring to the remote (also MDI) PHY, as it says, then row 1 assigns the signals BI_DA+ and BI_DB+ to contact 1. Can you please clarify with an example?
    – claymation
    Commented Jan 13 at 0:32
  • "No, that won't work." So does that mean that the BI_DA signal pair in one PHY is connected to the BI_DB signal pair in the link partner PHY?
    – claymation
    Commented Jan 13 at 0:33
  • Yes, exactly - that is what the table's accompanying text tries to say.
    – Zac67 ♦
    Commented Jan 13 at 7:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.